Dyna Robotics was founded by repeat entrepreneurs Lindon Gao and York Yang, who previously built and sold checkout-automation company Caper AI for roughly $350 million, together with former DeepMind research scientist Jason Ma. The company's thesis is that robotics has reached an inflection point where a single learned foundation model, rather than hand-engineered per-task software, can deliver commercial-grade performance across messy real-world environments.
Dyna's flagship system, DYNA-1, is a robotic foundation model that generalizes and self-improves across varied tasks and settings. In published testing the model ran fully autonomously for more than 24 hours, folding over 900 napkins at production quality with success rates above 99%. The company emphasizes 'commercial-grade' reliability as the bar that distinguishes deployable robots from research demos, and its dual-arm robots have been put to work in hotels, restaurants, laundromats and gyms operating long shifts each day.
The company first raised a $23.5 million seed round co-led by First Round Capital and CRV. In September 2025 it announced a $120 million Series A, with reporting citing leads including CRV and First Round alongside strategic investors such as Salesforce Ventures, NVentures (NVIDIA's venture arm), the Amazon Industrial Innovation Fund, Samsung Next and LG Technology Ventures. The new capital is earmarked for expanding the research and engineering team and accelerating development of Dyna's next-generation foundation model.
Dyna competes in the embodied-AI foundation-model race that includes Physical Intelligence, Skild AI and Genesis AI, but it differentiates by anchoring on near-term commercial deployments and revenue-generating tasks rather than pure research milestones. By prioritizing robots that can run reliably in customer environments today, Dyna argues it can build the real-world data flywheel needed to push toward broader physical AGI.